Route-Preview Adaptive Model Predictive Motion Cueing for Driving Simulators

Motion cueing algorithm (MCA) aims to reproduce the dynamic motion experience of real vehicles for users of driving simulators. Under rough or irregular road conditions, vehicles are subjected to severe shocks and vibrations. However, due to the inherent response delay and limited capability of motion platforms in reproducing high-frequency components, conventional MCA often suffers from slow response and poor tracking accuracy. This mismatch leads to dynamic inconsistency between the visual feedback and the motion cues provided to the driver, which can easily induce discomfort or even aggravate simulator sickness. To address these issues, this study proposes a route-preview MCA based on adaptive model predictive control (RPAMPC). A CNN–LSTM-based vehicle trajectory prediction model is developed by integrating convolutional and recurrent neural networks to exploit forward terrain information. Subsequently, a motion cueing prediction model incorporating actuator stroke and velocity states is formulated, and an AMPC-based MCA is designed to optimize the simulator platform motion under physical constraints. Experimental results on a Stewart motion simulation platform demonstrate that, compared with traditional MCA, the proposed algorithm achieves higher-quality motion cues and significantly reduces sensory errors under complex road conditions.